In recruitment, a generative answer does not merely summarize information; it silently constructs a decision surface.
What the phenomenon looks like
The model infers suitability criteria from job descriptions, employer signals, or public norms, then acts as if those criteria had been explicitly stated. What was only implicit in the discourse becomes a decision filter.
Why it happens
The model fills gaps by borrowing the nearest stable pattern from public discourse, documentation, and training priors. The result is often coherent, but coherence here comes from inference, not from authorized interpretation.
Why it matters
This is risky because undeclared criteria are hard to contest. They can introduce bias, narrow access, and produce discriminatory effects without any visible policy statement to challenge.
What must be governed
- Define admissible evaluation criteria explicitly instead of letting the model infer them from cultural priors.
- Separate descriptive role information from ranking logic and fit assessment.
- Audit for implicit exclusions whenever the answer starts to sort, filter, or prioritize candidates.
